Gradient based solver for Support Vector Machines (SVM)
Gradient based solver for SVM
This is a multi-class classifier. The classification was done on the Fashion-MNIST dataset, which has 10 classes.
The accuracy achieved was 80.2%. The confusion matrix is in the report. The sklearn-library achieved an accuracy 80.37% on the same dataset.
Python3, Linux
To run the code and replicate results:
Clone the fashion mnist repository using
git clone https://github.com/zalandoresearch/fashion-mnist.git
Copy the code (pegasosSVM.py) into the fashion mnist directory
Run “python3 pegasosSVM.py” in the terminal. It takes around 10-13 minutes to execute